Misinformation Span Detection - EI22 & BOL4Y

BOL4Y & EI22

Description of each data provided in the paper.

Contents

The repository includes the following data files:

- dump.csv — Parsed and normalized fact-check metadata extracted from AosFatos HTML pages

- aos_fatos_pages.zip — Raw HTML pages crawled from AosFatos (BOL4Y)

- escriba-csv.zip — Transcripts from escriba

- bol4y.csv.gz — Main BOL4Y dataset (compressed CSV)

- ei22.csv — Main EI22 dataset (CSV)

- bol4y-transcripts.zip — Transcripts associated with BOL4Y

- ei22-transcripts.csv — Transcripts associated with EI22

Data

dump.csv

This file contains parsed fact-checking metadata from AosFatos, used to build BOL4Y.Original data is in Brazilian Portuguese; column names were translated to English for usability.

Column name
Description

title
Title of the fact check

date
Publication date of the fact check

aos_fatos_link
URL of the original AosFatos page

fact_check
Fact-checking paragraph

topic_pt
Topic(s) of the false claim (Portuguese)

source
Origin of the claim (e.g., livestream, speech)

source_urls
URLs to the original content being fact-checked

repetition_count
Number of times the same claim was repeated

year_days_pair
Year, month, and day of each claim occurrence. Includes repeated claims

page
Crawl page index where the fact check was found

fact_check_id
Unique identifier for the fact check (shared with BOL4Y)

BOL4Y (bol4y.csv.gz)

CSV was compressed for easier storage. In it, you will find the following fields:

Column name
Description

fact_check_id
Identifier matching dump.csv

file
Source file for the transcript 

transcription_source
Transcription system (escriba or whisper)

transcription_index
Segment ID(s) associated with each segment. For some fact-checked segmentes, you'll see more than one id listed in this field, following the concatenation approach discussed in the paper

transcription_text
Claim or transcript segment text

label
1 = misinformation, 0 = non-misinformation

fixed_date
Date of the fact check

unique-check
Unique ID combining file and segment information

EI22

Each line of this csv is a segment

Column name
Description

file_id
Identifier of the source file

transcript_timestamp_ids
Segment ID(s) associated with each segment. For some fact-checked segmentes, you'll see more than one id listed in this field, following the concatenation approach discussed in the paper

transcript
Full transcript of the file (EI22 only)

minute_timestamp
Minute-level timestamp provided by AosFatos

description
Metadata description from AosFatos

label
1 = misinformation, 0 = non-misinformation

unique-check
Unique ID combining file and segment information

segment_text
Claim or segment text

Note: For youtube videos, the file_id field corresponds to the id of the video on youtube

Additional Resources
HuggingFace Repository

Github repository